*DO-FILE FOR "NATO SOURCES DATASET"
 
*This do-file corresponds with tables and figures featured in the article "Who Matters for Memory: Sources of Institutional Memory in International Organization Crisis Management" in the Review of International Organizations
*Tables and Figures indicated below are presented in the order in which they appear in the Main Manuscript. 
*The codebook for the dataset and a list of all variables in the dataset are located at the bottom of this file. 

**********MAIN MANUSCRIPT**********

*Table 2: Average Treatment Effects (ATEs)
*(a) IS Source Cue as Treatment
reg sccsr iscontrol
reg colleague iscontrol
reg sprvisr iscontrol

*(b) US Source Cue as Treatment
reg sccsr uscontrol
reg colleague uscontrol
reg sprvisr uscontrol

*Figure 2: Likelihood of Recording for Self or Successors by Institution
twoway scatter sccsr institution if treated==0, jitter(7) 
twoway scatter sccsr institution if treated==3, jitter(7) 

*Table 4-6
*Quotes from NATO elites

*APPENDIX
*Table 1: Subject Pool
*Indicates how many subjects sampled at each of six NATO institutions

*Figure 1: Recording for Self or Successors versus Discussing with Colleague
*Recording for Self or Successors
reg sccsr uscontrol
adjust, by(uscontrol) ci
*Copied these coefficients and confidence intervals into Excel and made the bar graphs

reg sccsr iscontrol
adjust, by(iscontrol) ci
*Copied these coefficients and confidence intervals into Excel and made the bar graphs

reg sccsr mediacontrol
adjust, by(mediacontrol) ci

reg colleague uscontrol
adjust, by(uscontrol) ci
*Copied coefficients and confidence intervals into Excel and made bar graphs

reg colleague iscontrol
adjust, by(iscontrol) ci
*Copied coefficients and confidence intervals into Excel and made bar graphs

reg colleague mediacontrol
adjust, by(mediacontrol) ci

*Table 3: Effect of US Source Cue on Recording the Error for Self or Successors
**Model 1 (just treatment)
ologit sccsr uscontrol 
**Model 2 (treatment with all controls)
ologit sccsr uscontrol institution age gender nationality hypsc workfor civmil fe currentyrs totalyrs

**********ONLINE RESOURCE**********

*TABLE OF CONTENTS
*I. THEORETICAL ARGUMENT
*Additional Discussion of Terminology

*II. HYPOTHESES
*Discussion of Hypothesis 4: The International Media

*III. METHODOLOGY
*Study Design Assumptions
*Subject Recruitment
*Dates and Locations of Experiment on Subjects
*Demographic and Institutional Characteristics
sum age gender nationality hypsc institution workfor civmil fe currentyrs totalyrs
*Block Randomization
*Pre-test Questions
*Survey Instrument as Treatment
*Outcome Questions

*IV. RESULTS
*1.	OVERVIEW OF TREATMENT EFFECTS
*Likelihood of Discussing with Supervisor
graph bar (mean) sprvisr, over(treated) blabel(bar)

*Successor: Differences in Means across Treatments versus Control
mean sccsr, over(treated)
oneway sccsr treated, tabulate

*Supervisor: Differences in Means across Treatments versus Control
mean sprvisr, over(treated)
oneway sprvisr treated, tabulate

*Colleague: Differences in Means across Treatments versus Control
mean colleague, over(treated)
oneway colleague treated, tabulate

*Joint Statistical Test – Seemingly Unrelated Regression with Supervisor
sureg (sccsr uscontrol)(colleague uscontrol)(sprvisr uscontrol)

*Cell Sizes across Source Cue Conditions on Recording for Self or Successors
tab sccsr treated

*Cell Sizes across Source Cue Conditions on Discussing with Supervisor
tab sprvisr treated

*Cell Sizes across Source Cue Conditions on Discussing with Colleague
tab colleague treated

*Balancing Statistics
mean institution if iscontrol==1
mean institution if uscontrol==1
mean institution if treated==0

mean age if iscontrol==1
mean age if uscontrol==1
mean age if treated==0

mean gender if iscontrol==1
mean gender if uscontrol==1
mean gender if treated==0

mean nationality if iscontrol==1
mean nationality if uscontrol==1
mean nationality if treated==0
 
mean hypsc if iscontrol==1
mean hypsc if uscontrol==1
mean hypsc if treated==0

mean workfor if iscontrol==1
mean workfor if uscontrol==1
mean workfor if treated==0

mean civmil if iscontrol==1
mean civmil if uscontrol==1
mean civmil  if treated==0

mean fe if iscontrol==1
mean fe if uscontrol==1
mean fe if treated==0

mean currentyrs if iscontrol==1
mean currentyrs if uscontrol==1
mean currentyrs if treated==0

mean totalyrs if iscontrol==1
mean totalyrs if uscontrol==1
mean totalyrs if treated==0

*2. INTERNATIONAL STAFF SOURCE CUE
*Effect of International Staff Cue on Recording for Self or Successors
**Logistic Model 3 (just treatment)
ologit sccsr iscontrol
**Logistic Model 4 (treatment with all controls)
ologit sccsr iscontrol institution isinstit age gender nationality hypsc workfor civmil fe currentyrs totalyrs
**OLS Model (just treatment)
regress sccsr iscontrol

*Effect of International Staff Cue on Discussing with Colleague
*Logistic Model (just treatment)
ologit colleague iscontrol
*Logistic Model (treatment with all controls) 
ologit colleague iscontrol institution isinstit age gender nationality hypsc workfor civmil fe currentyrs totalyrs
**OLS Model (just treatment)
regress colleague iscontrol

*Effect of International Staff Cue on Discussing with Supervisor
*OLS Model (just treatment)
regress sprvisr iscontrol

*3.	US SOURCE CUE
*US Source Cue on Recording for Self or Successors
** OLS Model (just treatment)
regress sccsr uscontrol 
** OLS Model 10 (treatment with all controls)
regress sccsr uscontrol institution age gender nationality hypsc workfor civmil fe currentyrs totalyrs

*US Source Cue on Discussing with a Colleague
**Logistic Model 5 (just treatment)
ologit colleague uscontrol 
**Logistic Model 6 (treatment with controls)
ologit colleague uscontrol institution age gender nationality hypsc workfor civmil fe currentyrs totalyrs
**OLS Model (just treatment)
regress colleague uscontrol

*US Source Cue on Discussing with a Supervisor
regress sprvisr uscontrol

*Effect of US Source Cue on American Elites
**1. I generated a new variable "american" indicating 1 if American, 0 if non-American
*gen american = 0
*replace american = 1 if nationality==28
**2. I created an interaction term for American receiving the US treatment
*gen american_uscontrol = american*uscontrol 
**3. I then regressed successor variable on US treatment and americans who received the US treatment
reg sccsr uscontrol american american_uscontrol
**4. I also regressed colleague variable on US treatment and americans who received the US treatment
reg colleague uscontrol american american_uscontrol

*4.	INTERNATIONAL MEDIA SOURCE CUE
*Effect of International Media Source Cue on Recording for Self or Successors
*Logistic Model (just treatment)
ologit sccsr mediacontrol
*Logistic Model (treatment with controls)
ologit sccsr mediacontrol institution age gender nationality hypsc workfor civmil fe currentyrs totalyrs
**OLS Model (just treatment)
regress sccsr mediacontrol

*Effect of International Media Source Cue on Discussing with a Colleague
*Logistic Model (just treatment)
ologit colleague mediacontrol
*Logistic Model (treatment with controls)
ologit colleague mediacontrol institution age gender nationality hypsc workfor civmil fe currentyrs totalyrs
**OLS Model (just treatment)
regress colleague mediacontrol

*Effect of International Media Source Cue on Discussing with a Supervisor
**OLS Model (just treatment)
regress sprvisr mediacontrol

*5. CORRELATIONS
**Correlations between US Source Cue and Independent Variables / Control Variables
corr uscontrol institution
corr uscontrol age
corr uscontrol gender
corr uscontrol nationality
corr uscontrol hypsc
corr uscontrol workfor
corr uscontrol civmil
corr uscontrol fe
corr uscontrol currentyrs
corr uscontrol totalyrs

**Correlations between IS Source Cue and Independent Variables / Control Variables
corr iscontrol sccsr
corr iscontrol colleague
corr iscontrol sprvisr
corr iscontrol institution
corr iscontrol age
corr iscontrol gender
corr iscontrol nationality
corr iscontrol hypsc
corr iscontrol workfor
corr iscontrol civmil
corr iscontrol fe
corr iscontrol currentyrs
corr iscontrol totalyrs

**Correlations between Media Source Cue and Independent Variables / Control Variables
corr mediacontrol sccsr
corr mediacontrol colleague
corr mediacontrol sprvisr
corr mediacontrol institution
corr mediacontrol age
corr mediacontrol gender
corr mediacontrol nationality
corr mediacontrol hypsc
corr mediacontrol workfor
corr mediacontrol civmil
corr mediacontrol fe
corr mediacontrol currentyrs
corr mediacontrol totalyrs

*Correlations between Sources and Control Variables
*Data from above all entered into Excel as a bar graph

*6. IRB Protocol as Pre-Registration

**********DATA*************
*The following variables are included in the dataset:
*iscontrol
*mediacontrol
*uscontrol
*colleague
*institution
*sccsr
*sprvisr
*treated
*isinstit

*The following demographic data has been removed from the dataset in order to protect the anonymity of human subjects, in accordance with IRB protocol, as anonymity was assured in exchange for subjects' participation.
*age
*currentyrs
*fe
*gender
*hypsc
*civmil
*nationality
*totalyrs
*workfor
*american
*american_uscontrol

**********CODEBOOK**********
*VARIABLES IN THE "NATO SOURCES DATASET"
*Variables correspond with outcome questions, treatments and pre-test questions as indicated in the Methodology section of the Online Resource

*DEPENDENT VARIABLES

*sccsr = likelihood of recording for self or successors
***Question: How likely would you be to record this for yourself or successors?
** 0 if Not at all likely
** .25 if Unlikely
** .5 if Somewhat likely
** .75 if Very Likely
** 1 if Definitely likely

*sprvisr = likelihood of discussing with supervisor
***Question: How likely would you be to discuss this with your supervisor?
** 0 if Not at all likely
** .25 if Unlikely
** .5 if Somewhat likely
** .75 if Very Likely
** 1 if Definitely likely

*colleague = likelihood of discussing with close colleague
***Question: How likely would you be to discuss this with a close colleague?
** 0 if Not at all likely
** .25 if Unlikely
** .5 if Somewhat likely
** .75 if Very Likely
** 1 if Definitely likely


*INDEPENDENT VARIABLES

*treated = indicator of whether elite received one of three treatments or a placebo condition
** 0 if Placebo
** 1 if International Staff
** 2 if International Media
** 3 if US

*iscontrol = International Staff treatment as source cue 
***Elite received a hypothetical scenario in which the International Staff was used as treatment
** 0 if elite did not receive IS treatment
** 1 if elite did receive IS treatment

*mediacontrol = international media treatment as a source cue
***Elite received a hypothetical scenario in which the International Staff was used as treatment
** 0 if elite did not receive media treatment
** 1 if elite did receive media treatment

*uscontrol = US treatment as a source cue
***Elite received a hypothetical scenario in which the International Staff was used as treatment
** 0 if elite did not receive US treatment
** 1 if elite did receive US treatment


*CONTROL VARIABLES

*institution = Elite's current institution at NATO where he/she is physically working - irrespective of his/her employer
**NOTE: For example, an elite may be working as a military staff member in IMS at NATO Headquarters but be employed by his/her member state. 
* 1	if Allied Command Operations (ACO)
* 2	if Allied Command Transformation (ACT)
* 3	if International Military Staff (IMS)
* 4	if International Staff (IS)
* 5	if Military Committee (MC)
* 6	if North Atlantic Council (NAC)

*isinstit = interaction term of NATO elites working for the IS and who also received the IS source cue
**Original code for creating this variable was the following: generate isinstit = iscontrol*institution

*civmil = Elite's job status as civilian or military
** 0 if civilian
** 1 if military
*Question: Are you considered civilian staff or military staff?

*workfor = Elite's employer
** 0 if a state (e.g. France)
** 1 if NATO
*Question: Are you employed by NATO, a nation or do you work as a VNC?
**NOTE: Ultimately, I did not recruit any voluntary national contribution (VNCs) for this study because they did not fit the requirements for subject recruitment as indicated in the paper. Therefore, there is no code listed here for VNC.

*currentyrs = Numerical value indicating the elite's current years working in the current position in NATO
*Question: How many years have you worked in this position in NATO?

*totalyrs = Numerical value for the elite's total years in his/her lifetime working in NATO
*Question: How many years have you worked in your lifetime in NATO?

*fe = Elite has had or not had field experience of at least three months in a NATO crisis management operation
** 0 if no
** 1 if yes
*Question: Have you had field experience of at least three months in a NATO crisis management operation?

*hypsc = Elite has had or not responded to a hypothetical scenario as part of a study before
** 0 if no
** 1 if yes
*Question: Prior to this, have you responded to a hypothetical scenario as part of a study before?

*nationality = Elite's nationality
* 1	 Albania
* 2	 Belgium
* 3	 Bulgaria
* 4	 Canada
* 5	 Croatia
* 6	 Czech Republic
* 7	 Denmark
* 8	 Estonia
* 9	 France
* 10 Germany
* 11 Greece
* 12 Hungary
* 13 Iceland
* 14 Italy
* 15 Latvia
* 16 Lithuania
* 17 Luxembourg
* 18 Netherlands
* 19 Norway
* 20 Poland
* 21 Portugal
* 22 Romania
* 23 Slovakia
* 24 Slovenia
* 25 Spain
* 26 Turkey
* 27 United Kingdom
* 28 United States
*Question: What is your nationality?

*gender = Elite's gender identity
** 0 if Male
** 1 if Female
*Question: What is your gender?

*age = Numerical variable to indicate elite's age
*Question: What is your age?

*american = Elite is American or not
** 0 if not American
** 1 if American















